Enhancing Anomaly Detection in Automated Guided Vehicles through Feature Weight Optimization

In this study, we propose an approach for optimizing feature weights to enhance the detection of mechanical wear in Automated Guided Vehicles (AGVs) based on power consumption data. Leveraging genetic algorithms, we aim to maximize forecasting quality on normal operational data while amplifying errors on anomalous fragments. Our method utilizes a custom telemetry dataset and Long Short-Term Memory (LSTM)-based forecasting models. Experimental results demonstrate the effectiveness of the proposed optimization method in improving anomaly detection performance, showcasing significant enhancements in precision, accuracy, and overall predictive maintenance capabilities for AGVs.